Performance Guide

How It Works

Each tool call flows through: MCP protocol → Python → COM → Excel → response. Individual COM calls take 10–50ms, so the key to performance is minimizing COM round-trips per tool invocation.

Best Practices

1. Start with Sheet Summary

Always call read_data() without cell_range first:

{ "sheet": "Sheet1" }

This returns dimensions, headers, merged cells, and regions with only ~5 COM calls — regardless of sheet size. Use the summary to decide which ranges to read next.

2. Use Batch Operations

Instead of...	Use...	Savings
5× read_data() per sheet	1× read_data(sheet="*")	5 calls → 1
N× read_data(detail=true) per cell	1× get_formulas(cell_range="A1:Z100")	N calls → 1
N× checking cell formatting	1× get_cell_styles(cell_range="...")	N calls → 1

3. Filter Style Properties

get_cell_styles makes multiple COM calls per cell. Use properties to query only what you need:

{ "cell_range": "A1:Z50", "properties": ["bold", "bg_color"] }

This skips font_name, font_size, alignment, border, etc. — significantly reducing COM calls.

4. Use merge_info Selectively

merge_info=true checks every cell in the range for merge status (~1 COM call per cell). For large ranges (1000+ cells), this can be slow. Use it on specific sections, not entire sheets.

5. Large File Strategy

For workbooks with thousands of rows:

1. read_data() → get sheet summary (dimensions, headers)
2. read_data(cell_range="A1:Z20") → read header area to understand structure
3. Decide on target ranges based on the structure
4. Read specific ranges only — avoid reading the entire sheet

Performance Reference

Operation	Typical Time	COM Calls
Sheet summary (read_data())	~50ms	~5
Read 100 rows (bulk)	~30ms	1
Read 1,000 rows (bulk)	~100ms	1
merge_info on 100 cells	~200ms	~100
get_formulas on 1,000 cells	~50ms	2
get_cell_styles on 100 cells	~500ms	~100 × N props
get_objects	~20ms	~3